Regional sources of atmospheric formaldehyde and acetaldehyde, and implications for atmospheric modeling

Atmospheric Environment 47 (2012) 477e490

Contents lists available at SciVerse ScienceDirect

Atmospheric Environment journal homepage: www.elsevier.com/locate/atmosenv

Regional sources of atmospheric formaldehyde and acetaldehyde, and implications for atmospheric modeling D.J. Luecken*, W.T. Hutzell, M.L. Strum, G.A. Pouliot U.S. Environmental Protection Agency, 109 T.W. Alexander Dr., Mail Drop E243-03, Research Triangle Park, NC 27709, USA

a r t i c l e i n f o

a b s t r a c t

Article history: Received 17 June 2011 Received in revised form 30 September 2011 Accepted 3 October 2011

Formaldehyde and acetaldehyde concentrations over the Eastern half of the United States are simulated with a 3-D air quality model to identify the most important chemical precursors under January and July conditions. We find that both aldehydes primarily result from photochemical production, although 25% or more result from direct emissions in urban areas during winter. Isoprene is the major precursor of formaldehyde in most areas during summer, contributing 20e60% of total production, with the magnitude being spatially variable. Other alkenes from anthropogenic and/or biogenic emissions dominate formaldehyde production in winter, contributing 60e85% of total formation, and are prominent contributors in summer. Alkenes, including biogenic alkenes, dominate acetaldehyde production during both seasons. These conclusions are based on the degradation of emitted VOCs described by the SAPRC07TB chemical mechanism, but even this detailed model has difficulty reproducing observed values better than a factor of 2. The substantial role of isoprene and other alkenes in aldehyde formation emphasizes that we examine and improve emission estimates of these compounds. Until we can estimate the emissions and understand the chemistry of VOC precursors to aldehyde formation with greater certainty, it will be difficult to accurately predict atmospheric concentrations of aldehydes and develop strategies to reduce their concentrations. Published by Elsevier Ltd.

Keywords: Formaldehyde Acetaldehyde HAPs Photochemical production Isoprene

1. Introduction Formaldehyde (HCHO) and acetaldehyde (CH3CHO) play important roles in atmospheric photochemistry and air quality. They are a large component of the total VOC reactivity of the atmosphere, providing sources of new radicals that drive ozone formation (Jeffries, 1995; Steiner et al., 2008). They are hazardous air pollutants (HAPs) known or suspected as carcinogens posing health risks (HEI, 2007; Tilgner and Herrmann, 2010; US EPA, 2011). They participate in aqueous chemistry due to their high solubility and interactions with other dissolved species in clouds (Adewuyi et al., 1984; Tilgner and Herrmann, 2010; Warneck, 1989), and contribute to secondary organic aerosols by producing radicals that oxidize hydrocarbons. The chemistry of these two compounds is an indicator of our overall understanding of atmospheric photochemistry. A better understanding of aldehyde production and decay will enable us to develop and analyze policies that consider the effects of air quality measures on a number of different atmospheric pollutants. Both compounds are emitted directly into the atmosphere from a variety of sources. Anthropogenic emissions of HCHO and CH3CHO * Corresponding author. Tel.: þ1 919 541 02544; fax: þ1 919 685 3236. E-mail address: [email protected] (D.J. Luecken). 1352-2310/$ e see front matter Published by Elsevier Ltd. doi:10.1016/j.atmosenv.2011.10.005

are ubiquitous, concentrated in urban areas, and are among the 25 highest-emitted anthropogenic VOCs in the US (Simon et al., 2010). Mobile sources and stationary combustion dominate anthropogenic sources. Biogenic sources include live and decaying plants, biomass burning, and seawater. Another source of HCHO and CH3CHO is secondary formation from degradation of VOCs (Atkinson, 1997; Seinfeld and Pandis, 1998), produced in multiple steps during hydrocarbon oxidation. Equations (1a) through (1c) show a simple example of HCHO formation from methane (CH4). Hydrogen abstraction from CH4, followed by fast addition of oxygen to the alkyl radical (1a) creates a methyl peroxy radical which can oxidize NO to NO2 (1b), creating a methyl alkoxy radical and eventually leading to ozone formation. A hydrogen atom from the alkoxy radical is abstracted during the reaction with oxygen (1c) along with elimination of HO2 to form formaldehyde: CH4 þ OH þ O2 / CH3O2 þ H2O

(1a)

CH3O2 þ NO / CH3O þ NO2

(1b)

CH3O þ O2 / HCHO þ HO2

(1c)









Aldehydes can also be formed from alkoxy radicals after scission of a CeC bond, through photolysis, and through alkoxy radical

478

D.J. Luecken et al. / Atmospheric Environment 47 (2012) 477e490

formation in the absence of nitrogen oxides (NOx), when reactions of the peroxy radical with HO2 and other peroxy radicals dominate. Because HCHO and CH3CHO sources include direct emissions and photochemical production, both sources must be correctly characterized when predicting atmospheric concentrations. The importance of emissions versus photochemistry depends on location, time of day and season, and affects any conclusions about how emission changes might be expected to change aldehydes. An earlier modeling study (Luecken et al., 2006) determined that over 80% of formaldehyde and acetaldehyde was due to secondary production. In urban areas in winter, emissions contributed up to 50%. Statistical techniques have been used with measurements to separate out the two components, concluding that 42% of HCHO is due to primary emissions in Mexico City (Garcia et al., 2006), 44e95% in Los Angeles (Grosjean et al., 1983), and approximately 33%e47% in Houston (Friedfeld et al., 2002; Rappenglück et al., 2010). These measurements represent a “true” value that models must attempt to match, but they are made in cities with large emission sources, and include uncertainties such as unspeciated background contributions. Measurements from a few sites may not represent atmospheric processes at other places and times. Both aldehydes are included in atmospheric chemical mechanisms although their representation varies among mechanisms. A large number of organic compounds participate in atmospheric chemistry and each cannot be explicitly represented in the sequence of secondary, tertiary, quaternary, etc. reactions that result in the production of aldehydes, especially at regional-to-global scales. The Master Chemical Mechanism (MCM) (Saunders et al., 2003) contains over 12,600 reactions for 4500 species, but its size makes it unsuitable for repeated simulations across a large area. Chemical mechanisms in 3-D air quality models (AQMs) are significantly condensed, invoking various assumptions and simplifications in the chemistry. In this study, we use the Community Multiscale Air Quality (CMAQ) model (Byun and Schere, 2006) with the SAPRC07TB chemical mechanism (Hutzell et al., 2012) to track sources of ambient HCHO and CH3CHO. A recent evaluation of CMAQ with SAPRC07TB has been made for ozone and fine particulate matter (PM2.5) components. With this system, we simulate detailed atmospheric photochemical formation and decay of HCHO and CH3CHO in both low- and high-NOx conditions, within a realistic

environment that is affected by atmospheric transport, deposition, precipitation, actinic flux, temperature and pressure gradients, and spatially- and temporally-varying emissions. An increasingly important use of CMAQ is to assess proposed emission changes on a multipollutant basis, where effects of emission changes on ambient concentrations of ozone, PM2.5, and hazardous air pollutants are predicted simultaneously. The goal of this work is to characterize regional-scale concentrations and chemical sources of atmospheric HCHO and CH3CHO, and identify which VOCs are the largest producers of these aldehydes. Our analysis differs from previous studies by 1.) explicitly tracking direct emissions and photochemical production, 2.) employing a smaller grid resolution than previous aldehyde modeling studies, 3.) allocating production to specific, emitted compounds, and 4.) examining spatial and temporal differences in contributions. The ability to identify direct and indirect contributions to atmospheric HCHO and CH3CHO will help us predict how they will respond to emission changes, and guide development of policies that reduce ozone and PM2.5 without increasing HAPs. This can also assist in prioritizing research efforts to improve aldehyde predictions, by focusing on those processes likely to have the largest impact on concentrations. 2. Methods The model and methods have been previously described (Hutzell et al., 2012) and are briefly summarized here. 2.1. CMAQ model, domain, boundary conditions and meteorology We applied CMAQ version 4.7.1 (Foley et al., 2009) with a 36  36 km2 horizontal grid covering the continental US and a nested 12  12 km2 grid over the Eastern half of the US (Fig. 1). There are 14 vertical layers, extending from the surface up to about 16 km. Boundary conditions for the continental simulations were extracted from the GEOS-CHEM model with those for the nested domain provided by continental CMAQ simulations. Meteorological inputs were obtained from the MM5 meteorological model (http:// box.mmm.ucar.edu/mm5/) version 3.6.1. Simulations were performed for two months, January and July, 2002, after a10-day model spin up period.

50

Minneapolis

45

Chicago New York E. Penn Baltimore Louisville Atlanta* N. Georgia

40

35

Alabama Houston

30

25 -120

-110

-100

-90

-80

-70

Fig. 1. Modeled 36 km and 12 km domains. The locations of ten sites that are used in additional analyses are noted. Sites with an urban composition are shown by x and sites with large biogenic influences by .

D.J. Luecken et al. / Atmospheric Environment 47 (2012) 477e490

We created emissions using the 2002 National Emissions Inventory (NEI) (http://www.epa.gov/ttn/chief/net/2002inventory. html). The Sparse Matrix Operator Kernel Emissions (SMOKE) emission processor version 2.4 (http://www.smoke-model.org) was used to create model ready emissions, with speciation rules (http://www.cert.ucr.edu/wcarter/emitdb/) for allocating explicit species to lumped model species. Biogenic emissions were calculated with Biogenic Emissions Inventory System (BEIS) version 3.13. The 2002 HAP inventory was used to estimate HAP emissions and other VOCs were derived from speciation of total VOC emissions (Simon et al., 2010). 2.2. Chemical mechanism The SAPRC07 mechanism has been extensively evaluated and described in detail elsewhere (Cai et al., 2011; Carter, 2010). This study uses the SAPRC07TB version, which includes additional explicit high emission and HAP species, including propene, ethanol, 1,2,4-trimethylbenzene, acrolein, 1,3-butadiene, toluene, three isomers of xylene, and a-pinene. The CMAQ implementation of SAPRC07TB (Hutzell et al., 2012) includes temperature and pressure dependencies of the photolysis parameters and extensions for aerosol formation. 2.3. Process and integrated reaction rate analyses To quantify contributors to aldehyde formation, we implement the Integrated Reaction Rate (IRR) and Process Analysis (PA) capabilities within CMAQ (Gipson, 1999). This technique has been useful for understanding ozone production and cycling of radicals and nitrogen oxides (Jang et al., 1995; Jeffries and Tonnesen, 1994). We use IRR to track radical and oxidized intermediates which form HCHO and CH3CHO, such as peroxy and acyl peroxy radicals, methyl vinyl ketone and methyl ethyl ketone and allocate these intermediates to stable species which are emitted. Products from all VOC reactions are tracked if they eventually react to form either HCHO or CH3CHO. The entire chemical mechanism is utilized for the CMAQ simulation, and we use IRR within the simulation to quantify the HCHO and/or CH3CHO produced through 251 organic reactions (Supplementary Information, Table S-1). This information is postprocessed to quantify HCHO and/or CH3CHO produced from emitted organic species. Since some VOCs are both emitted and produced photochemically, we separate the emitted portion from that formed photochemically, which we track to its precursors. For example, the model species representing methyl ethyl ketone and similar ketones (MEK), an emitted VOC, produces HCHO via acetyl and methyl peroxy radicals. But MEK can itself be produced by isoprene as a second generation product. We calculate the fraction of MEK concentration at each hour due to emissions, which ranges from 10e95%, depending on location, and use that to apportion HCHO to either MEK emissions or its precursors (in this example, isoprene). Similarly, we separate the formation of HCHO from CH3CHO into that fraction due to emissions and to chemistry. We assume that model species transported into the grid cell have the same ratio of formation versus emission as the current grid cell. We list results for four explicit species (isoprene, methane, acetaldehyde and ethanol) and 9 lumped chemical categories:  alkanes: non-methane alkanes lumped in model species ALK1 through ALK5.  ethene and propene.  lumped alkenes: model species OLE1 and OLE2.  aromatics: benzene, m-,p- and o-xylene, toluene, 1,2,4trimethylbenzene, and model species ARO1 and ARO2.

479

 terpenes and pinenes: a-pinene and model species SESQ and TERP, representing sesquiterpenes and other terpenes.  PANs: peroxyacetyl nitrate, peroxymethacryloyl nitrate, and model species PAN2 (peroxypriopionyl nitrate and higher PANs) and PBZN (from aromatic aldehydes).  C3þ aldehydes: model species RCHO representing aldehydes with 3 or more carbons.  lumped oxygenates: methylglyoxal, acetone, and model species MEK, lumped ketones (PRD2) and isoprene products (IPRD).  other: species that individually produce <1% of total formation. We selected 10 locations throughout the domain, shown in Fig. 1, for performing IRR analyses, to examine how different chemical regimes and emissions mixes affect aldehyde precursors. These locations include five grid cells examined previously (Hutzell et al., 2012) for their oxidant chemistry (Louisville, Baltimore, Atlanta, Alabama, N. Georgia), three rural locations (Alabama, E. Pennsylvania, N. Georgia), and five urban locations where measurements are available in 2002 for comparison (Chicago, Baltimore, New York, Houston, Minneapolis). These locations, defined by processes within a 12  12 km2 grid cell at the given location, are not meant to fully characterize specific cities or regions, but do display the range of variability expected in contributors to aldehyde formation. 3. Results and discussion 3.1. Regional concentrations and impact of direct emissions versus photochemical production We focus on results over the Eastern nested domain because the finer scale more accurately characterizes spatial variability in meteorological and photochemical processes. Fig. 2 shows monthly-averaged concentrations of HCHO in January and July, and fraction of overall concentration due to photochemistry. The remainder is attributed to direct HCHO emissions. In Fig. 2, modeled HCHO concentrations are higher in July than January, by over a factor of 4, due to higher actinic flux, radical concentrations and precursor emissions of biogenic VOCs in summer. During winter, concentrations are highest in urban areas where emissions of HCHO are largest; in these areas emissions account for 25e50% of total HCHO. During summer, HCHO emissions contribute much less: 6e25%. The photochemical contribution and concentration of HCHO is highest in the Southeast, consistent with satellite observations (Millet et al., 2008; Palmer et al., 2006) and earlier modeling results (Luecken et al., 2006). Fig. 3 shows concentrations and emission fractions for CH3CHO. The seasonal patterns are similar to HCHO, but concentrations are lower by half and the contribution of photochemistry is greater, accounting for over 80% of total CH3CHO in winter and 95% in summer. Fig. 4 displays seasonal differences in the role of emissions versus production at the 10, mostly urban locations, and the average over the entire domain. For this sampling of grid cells, primary emissions contribute at most 55% of the HCHO in January (Louisville). In summer, emissions contribute less than 20% at all locations. 3.2. Chemical precursors of aldehydes Relative contributions of different VOCs to aldehyde production depend on the magnitude of emissions and yield of aldehydes per molecule of reacting VOC. We use IRR and PA analyses to quantify the role of different VOCs by summing reaction rates and yields over two periods: January 14e16, 2002 and July 15e17, 2002, focusing on 10 locations.

480

D.J. Luecken et al. / Atmospheric Environment 47 (2012) 477e490

Fractional HCHO formation attributed to individual VOCs and lumped classes is shown in Fig. 5. In general, isoprene is the dominant precursor of HCHO during July, although the contribution varies spatially. In the Southeast, such as the location near urban Atlanta, isoprene contributes approximately 50e65% of total formation. In the Mid Atlantic location near Baltimore, isoprene contributes approximately 20% of HCHO formation. The reaction of isoprene with OH dominates HCHO formation when isoprene is present, accounting for 75e95% of HCHO attributed to isoprene, largely through an operator in SAPRC07TB (xHCHO) which yields 80e99% HCHO in later sequences. Another operator produced by isoprene (xIPRD), produces reactive oxygenated products (IPRD), providing an additional 3e8% HCHO formation. During January, when emissions of isoprene are reduced, production is dominated by lumped and explicit alkenes. In the locations near New York and Louisville, explicit and lumped alkenes are larger contributors than isoprene, even in July. Methane and other alkanes also contribute to HCHO formation.

Fig. 6 displays VOC contributors to CH3CHO production. Isoprene plays a minor role in CH3CHO production, and alkenes contribute the largest amount. Terpenes and cyclic terpenes are major contributors in Southeastern locations. Alkanes and PANs, particularly peroxyproprional nitrate, contribute in some locations. Ethanol contributes up to 5% of CH3CHO production, and indirectly produces HCHO through CH3CHO decay. While lumped alkenes are the largest relative contributors in January, the absolute contribution of these compounds to HCHO and CH3CHO formation is smaller than in July. Both compounds are formed in larger amounts in summer (20e50 times higher for HCHO and 7e40 times higher for CH3CHO) because the major reaction involves the hydroxyl radical (OH), which can be 5 times or more higher in summer than winter (Ren et al., 2006). Supplementary Fig. S-1 shows a comparison of the total production of HCHO and CH3CHO from lumped alkenes for both periods. Additional factors, such as warmer temperatures, higher actinic fluxes and greater biogenic emissions increase formation in summer. The same

Fig. 2. Average HCHO concentration (left) and fraction of total HCHO due to photochemical production (right) in January (top) and July (bottom). Concentrations are in mg m3 and are averaged over the entire month.

D.J. Luecken et al. / Atmospheric Environment 47 (2012) 477e490

481

Fig. 3. Average CH3CHO concentration (left) and fraction of total due to photochemical production (right) in January (top) and July (bottom). Concentrations are in mg m3 and are averaged over the entire month.

seasonal factors that cause higher production rates via OH also cause higher destruction through OH, so concentration differences between summer and winter are not as large as suggested by formation rate differences (Figs. 2 and 3). 3.3. Comparisons of model results with observations Fig. 7 compares modeled HCHO and CH3CHO with 1, 3 and 24 haveraged observations from the PAMs network (www.epa.gov/air/ oaqps/pams), where the sites are differentiated by state. The observations are limited and insufficient for fully evaluating the model, but do serve as a touchstone to determine whether model predictions are realistic. One limitation of these observations is that sampling frequency varies: some sites report every day, others report every 3e12 days or longer, sometimes only one measurement is reported during the month. In addition, the monitoring sites are sparse and unevenly distributed spatially, making it difficult to evaluate predictions from a regional model. In July, 1299

observations from 96 sites were available, versus over 41 million model surface predictions, so we are evaluating only a small fraction of predicted values. In January, only 79 sites and 331 observations were available. Observations are particularly lacking in rural areas and the Southeast where high isoprene emissions and HCHO concentrations are predicted. In comparison with these limited observations, both compounds are underestimated by the CMAQ simulations, especially at higher concentrations. In July, the mean bias (MB) is 4 ppb, and normalized mean bias (NMB) is 0.59. When the highest 5% of observed values are removed, some of which are extremely high, the MB is reduced to 2.43 and NMB to 0.47. In January, the MB is 1.5 and NMB is 0.69, while with the top 5% of values removed, the MB is 0.83 and NMB is 0.55. In many states the bias is much less. It is unclear why some states are modeled well but others underpredicted. At some sites, such as those located around Chicago (IL), and Minneapolis (MN), the model matches predictions well, while others, including New York (NY) and RI are

482

D.J. Luecken et al. / Atmospheric Environment 47 (2012) 477e490

CH3CHO, Jan CH3CHO, July

HCHO, Jan HCHO, July

Fractional contribution from emissions

1

1

0.8

0.8

0.6

0.6

0.4

0.4

0.2

0.2

0

0 New

YorkaltimoreouisvilleChicagoHouston Minn AtlantaAlabamaE.Penn .Georgiamean L B N

Sample grid cell Fig. 4. Average fraction of HCHO and CH3CHO due to emissions at each of the 10 locations and the mean over the entire domain. Values are averaged over 3 days in January (solid bars) and July (hatched bars) at the 12 km grid cells sited at each of the locations.

modeled 50% too low. Others, including the Baltimore site (MD) fall in between. Previous attempts to model HCHO had limited success; sometimes models overpredict, but more often underpredict measured values at the surface. Previous comparisons of CMAQ predictions for HAPs (Luecken et al., 2006) showed that observations could generally be predicted within a factor or two. Comparisons of a box model and GEOS-CHEM predictions against data from the 2004 INTEX campaign (Fried et al., 2008) similarly showed 50% underpredictions. Several possible causes of differences between models and measurements have been noted previously. One cause is error in the measurements. Numerous studies have identified interferences in measurements of HCHO and CH3CHO (Arnts and Tejada, 1989; Herrington et al., 2006; Karst et al., 1993; Sirju and Shepson, 1995), which can cause positive or negative biases. Another possible cause is mismatch between point measurements and volume average concentrations predicted by models. The presence of large, localized sources may explain the inability of a 12  12 km2 grid cell to capture the highest values. In a few cases, these values are abnormally high: one site in Michigan reports a 24-h average HCHO concentration on July 13, 2002 of 71.1 mg m3, while the next measurement on July 25 is 4.81 mg m3. This monitor is south of downtown Detroit, in an area with industrial units, residential housing and highway traffic, so it could be influenced by intermittent high point emissions. A nearby monitor reported 5.79 mg m3 on July 13. Other errors could occur in specification of emissions or chemistry. To identify the largest potential sources of error causing underpredictions, we use information from Section 3 to focus on those chemicals with the largest influence on aldehyde predictions. We discuss the potential magnitude of uncertainties in emissions and chemistry of these VOCs, to prioritize areas for improving aldehyde predictions.

4. Implications for modeling the atmospheric chemistry of aldehydes Our ability to accurately predict ambient concentrations of formaldehyde and acetaldehyde depends on our ability to predict

concentrations of all VOCs. In this section, we examine factors influencing the concentrations of the most important VOC precursors identified in Section 3.2. We first discuss uncertainties involved in modeling emissions, and then uncertainty in modeling the chemistry of these compounds. 4.1. Aldehyde emissions and decay Emissions of HCHO and CH3CHO can be important sources of ambient concentrations, but atmospheric production is generally larger (Figs. 2 and 3). For both aldehydes, emissions derived from the NEI are less than emissions derived from total VOC emissions applying speciation profiles (Simon et al., 2010). Differences between the two values may be caused by lack of reporting requirements, different emission factors and speciation profiles. In the most extreme case (Louisville in January), using speciation profiles instead of the NEI could increase our modeled HCHO concentrations by a maximum of 50% and CH3CHO by 30% in winter, and HCHO up to 20% in summer, reducing some underprediction. In other areas, especially in summer, the change would only be a few percent. This suggests that direct emissions of formaldehyde and acetaldehyde are not likely to be the only cause of model underprediction. Emissions of larger aldehydes and subsequent photochemical decay are also a source of smaller aldehydes, and errors in these emissions could also increase the HCHO up to 24% and CH3CHO up to 29%. The photochemical loss reactions of HCHO and CH3CHO have been well studied and are not likely to have major errors. Slightly more than half of the photochemical loss is through photolysis, with most of the remainder from reaction with OH. 4.2. Isoprene Isoprene has been identified as a major contributor to HCHO in summer (Carslaw et al., 2000; Millet et al., 2006; Sumner et al., 2001), consistent with our study. Our results also show a substantial degree of spatial variability across the region, with contributions less than 20% in some areas. In rural locations, isoprene contributes about half of the HCHO production in summer. The ability to accurately predict HCHO concentrations in summer

D.J. Luecken et al. / Atmospheric Environment 47 (2012) 477e490

Fraction of total HCHO formation

0.7

0.7

Jan 0.6

0.6

0.5

0.5

0.4

0.4

0.3

0.3

0.2

0.2

0.1

0.1

0

0 York altimore ouisvilleChicagoHouston Minn L B New

a ia ta nn Atlan Alabam E.Pe N.Georg

0.7

Fraction of total HCHO formation

483

0.7

July 0.6

0.6

0.5

0.5

0.4

0.4

0.3

0.3

0.2

0.2

0.1

0.1

0

0 York altimore ouisvilleChicagoHouston Minn L B New

isoprene lumped alkenes (OLE1/2) ethene and propene methane

a ia ta nn Atlan Alabam E.Pe N.Georg terpenes and pinenes acetaldehyde (emis) PANs ethanol alkanes lumped oxygenates aromatics other

Fig. 5. Fraction of total formaldehyde attributed to primary VOC classes in January (top) and July (bottom) at 10 locations. Values are calculated from the sum over 72 h at the 12 km grid cells sited at each location.

strongly depends on accurately predicting isoprene concentrations, which depends on accurately representing both isoprene emissions and chemistry. Isoprene is almost entirely emitted by biogenic sources and estimates of the source strength are uncertain over regional scales. Several models of isoprene emissions are available, but they give different estimates. A comparison of four models showed a factor of two difference among models (Warneke et al., 2010). In modeling studies using BEIS, isoprene concentrations are underestimated (Doraiswamy et al., 2009), overestimated (Wiedinmyer et al., 2005) and reasonably well estimated (Hogrefe et al., 2011). Table 1 shows examples of differences that might be seen between inventories developed using two different models for biogenic emissions; the BEIS and MEGAN (Guenther et al., 2006) models, at each of the 10 locations. Isoprene emissions estimated by these two models can differ by 75% or more, with emissions from MEGAN generally larger than BEIS. A factor of 2 error in isoprene emissions could increase modeled formaldehyde by up to 80%. The large influence of isoprene on HCHO combined with large uncertainties in the

emissions of isoprene introduces significant uncertainties in HCHO predictions. The atmospheric chemistry of isoprene is currently being reexamined, and it is unclear how updated isoprene chemistry will affect predictions of HCHO. Inclusion of the 1,5-H isomerization reactions of b-hydroxyalkenyl peroxy radicals increased OH, and this channel would increase the HCHO production proportionally (Archibald et al., 2010a). Similarly, consideration of the 1,6-H shift leads to hydroperoxy aldehydes, yielding additional HCHO after rapid photolysis, along with OH and HO2 (Peeters, 2010; Stavrakou et al., 2010). Increased OH radical regeneration in nonurban environments could increase production of HCHO through other VOCs. Updated isoprene chemical reactions can have significantly higher HCHO yields from d-hydroxy isoprene nitrates (Paulot et al., 2009), and using this chemistry in CMAQ can potentially increase modeled HCHO less than 10% overall (Xie et al., 2011b). A critical evaluation of reaction products must be performed for any new isoprene mechanism. Because no consensus has yet been reached on the best way to represent isoprene chemistry, we offer no

484

D.J. Luecken et al. / Atmospheric Environment 47 (2012) 477e490

Fraction of total CH3CHO formation

0.7

0.7

Jan

0.6

0.6

0.5

0.5

0.4

0.4

0.3

0.3

0.2

0.2

0.1

0.1

0

0

York altimore ouisvilleChicagoHouston Minn L B New

a ia ta nn Atlan Alabam E.Pe N.Georg

Fraction of total CH3CHO formation

0.7

0.7

July 0.6

0.6

0.5

0.5

0.4

0.4

0.3

0.3

0.2

0.2

0.1

0.1

0

0

York altimore ouisvilleChicagoHouston Minn L B New isoprene lumped alkenes (OLE1/2) ethene and propene terpenes and pinenes

a ia ta nn Atlan Alabam E.Pe N.Georg

PANs alkanes aromatics C3+ aldehyde (emis)

ethanol lumped oxygenates other

Fig. 6. Fraction of total acetaldehyde attributed to primary VOC classes in January (top) and July (bottom) at 10 locations. Values are calculated from the sum over 72 h at the 12 km grid cells sited at each location.

recommendation here on improvements, but do expect it will affect HCHO. 4.3. a-pinene, terpenes and sesquiterpenes The terpenes and cyclic terpenes are known contributors to SOA formation, and this analysis shows that they also contribute to aldehyde formation. Acetaldehyde (up to 25%) and formaldehyde (6%) production depends on correctly estimating emissions of terpenes. These results agree with a previous analysis over Europe (Dufour et al., 2009) in which terpene emissions accounted for 8% of average CH3CHO in Western Europe, and up to 25% locally. Like isoprene, these chemicals are almost entirely biogenic but their emissions have not been studied in as much detail. We conclude that a similar uncertainty as isoprene is likely; uncertainty of a factor of 2 could increase acetaldehyde concentrations by 25%. The chemistry of terpenes beyond first generation products is also not well known, and many assumptions were made in its condensed representation. A comparison of SAPRC07 with the

MCM, which has a more complete description of intermediate chemistry, found that MCM predicted 2.3 times the photochemical ozone creation potential (POCP) for a-pinene than SAPRC07, almost twice as much for b-pinene, and 2.8 times more than limonene. This was attributed to different treatments of degradation chemistry beyond first generation products (Derwent et al., 2010). While ozone reactivity is not the same as aldehyde formation, they both occur through similar pathways and it is possible that aldehyde formation would be similarly lower in SAPRC07TB. If the mechanistic reactivity of the terpenes is underrepresented in SAPRC07TB by a factor of 2, the CH3CHO might be expected to increase up to 25%. 4.4. Other alkenes Other alkenes, including ethene, propene, and larger alkenes and dienes, are major contributors to both HCHO and CH3CHO formation, especially in the January simulation. In a previous allocation of acetaldehyde sources (Millet et al., 2010), alkenes

D.J. Luecken et al. / Atmospheric Environment 47 (2012) 477e490

10

1:1 line

Modeled HCHO, ppb

8

6 2:1 line

4

2

0 0

2

4 6 Observed HCHO, ppb

8

10

CT FL GA IL IN IA LA ME MD MA MI MN MS MO NE NH NJ NY OH PA RI SC TN TX VA VT WI

10

FL GA IL IN IA LA ME MD MA MI MN MS MO NE NH NJ NY OH PA RI SC TN TX VA VT WI

10

10

1:1 line

Modeled CH3CHO, ppb

8

6 2:1 line

4

2

0 0

2

4 6 Observed CH3CHO, ppb

8

1:1 line

8

6 2:1 line

4

2

0 0

2

4 6 Observed HCHO, ppb

8

10

1:1 line

8

6 2:1 line

4

2

0 0

10

485

2

4 6 Observed CH3CHO, ppb

8

10

Fig. 7. Comparison of observed and modeled HCHO in January (top, left) and July (top, right) and the same comparison for CH3CHO in January (left) and July (right). Different colors represent different States, which may have multiple monitors, and symbols represent whether sampling is 1 or 3 h intervals (o) or 24-h intervals (þ). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Table 1 Example of differences in emissions for alkene species predicted by two different models, BEIS and MEGAN. The emissions are in moles/sec and are the sum over 1 day on July 16, 2002. Location

New York Baltimore Louisville Chicago Houston Minn Atlanta Alabama E. Penn N. Georgia

Ethene þ Propene

Isoprene

OLE1 þ OLE2

BEIS

MEGAN

BEIS

MEGAN

BEIS

MEGAN

4.4 3.5 4.5 19.4 2.9 15.9 19.1 15.8 15.5 20.5

33.6 57.5 12.8 50.3 22.0 16.1 26.8 53.8 41.1 34.2

1

2.8 5.1 4.4 5.2 2.9 3.6 3.7 3.9 3.4 3.1

1.2 1 1.5 1.6 1.6 2.2 2.0 3.9 1.5 1.2

.4 .9 .7 .8 .8 .6 .9 1.4 .6 .6

.6 1.2 1 1.3 2 2.3 2.3 1.5 1.5

486

D.J. Luecken et al. / Atmospheric Environment 47 (2012) 477e490

Table 2 Fraction of total emissions of alkenes from biogenic sources for January and July simulations, based on the sum over one day on January 15, 2006 and one day on July 16, 2002. Location

New York Baltimore Louisville Chicago Houston Minn Atlanta Alabama E. Penn N. Georgia

Ethene þ Propene

OLE1 þ OLE2

Pinenes and terpenes

January

July

January

July

January

July

.00 .00 .00 .00 .12 .03 .15 .07 .07 .20

.08 .06 .16 .08 .35 .70 .59 .82 .79 .83

.00 .00 .01 .00 .10 .03 .14 .45 .09 .21

.08 .14 .24 .23 .31 .69 .53 .91 .83 .80

.02 .02 .04 .01 .55 .33 .86 .99 .29 .98

.52 .40 .65 .34 .82 .93 .98 1.00 .94 1.00

were about half as important as alkanes and slightly larger than ethanol. This contrasts with our findings, where alkenes contribute the majority of acetaldehyde production, perhaps due to inclusion of biogenic emissions. Many biogenic alkenes that are not isoprene or terpenes, such as 2-methyl-3-butene-2-ol, are lumped in OLE1 and OLE2. Table 2 lists the fraction of lumped alkenes from biogenic sources at each of the 10 locations (Supplementary Fig. S-2 shows the fraction across the Eastern US). In contrast to isoprene, alkenes from MEGAN are generally lower than BEIS for most locations, so using MEGAN would lower aldehyde production through alkenes. Are there reactive VOCs not predicted by MEGAN or BEIS that might contribute to HCHO formation? Photochemical modeling at Blodgett Forest (Choi et al., 2010) could not reconcile measured and modeled HCHO concentrations, postulating a missing reactive biogenic VOC source. Other analyses (Kim et al., 2011) found no evidence for significant missing biogenic VOCs in a branch enclosure measuring OH reactivity, and attribute discrepancies to unmeasured isoprene oxidation products. It is unclear whether errors in biogenic alkene emissions are a major source of aldehyde model underpredictions. Anthropogenic sources of lumped alkenes have equal or greater importance for aldehyde formation at all locations in winter, and urban locations in summer. Studies in Houston, Texas (Wert et al., 2003) agree with our study by pointing to anthropogenic emissions of propene and ethene as major HCHO precursors. Fig. 8 shows fifteen alkenes with the highest emissions and general source categories from which they are emitted (Simon et al., 2010). Mobile sources are the largest source of alkenes, along with emissions from general industrial sources, stationary combustion, and waste disposal, treatment and recovery. Mobile source estimation methods rely on models that account for emission differences due to fuel type, vehicle technology, speed, temperature, and modes of operation. There are uncertainties in the variability of vehicle fleets and fuels across the country, and in parameters used to convert monthly modeled emissions to hourly gridded emissions needed by models. For mobile sources, VOC speciation is generally well-characterized, and comparisons for HAP species such as HCHO and CH3CHO, which are reported both in the NEI inventory and computed from VOC speciation, are generally within 25% (Simon et al., 2010). Comparisons between emissions from the current MOBILE6 emissions model and the newer MOVES model show that mobile source VOC emissions from MOVES are similar to those from MOBILE6, sometimes slightly lower (Cook et al., 2010; Houyoux et al., 2009), or slightly higher (Kite, 2011). It is difficult to evaluate integrated emission inventories using ambient observations because measured concentrations include effects of meteorology, deposition, chemistry and emissions. In

a comparison of measurements in the Pacific Northwest with CMAQ modeling, the VOC reactivity of propene was overpredicted; reduction of the VOC inventory by 30% was needed for the model to match afternoon observations of ozone production efficiency (Xie et al., 2011a). Similarly, a modeling study using CMAQ in the Northeastern US found that modeled ethene concentrations generally overpredicted morning measurements by 20e100% (Doraiswamy et al., 2009). In contrast, an analysis using emission ratios found that inventories of alkenes in the NEI were underpredicted (Warneke et al., 2007). It is unclear whether anthropogenic alkene emissions in the NEI are too high or low; while errors in the inventory might increase HCHO and CH3CHO by 22 and 24% in the winter, they might also decrease predictions. Beyond the simplest alkenes, chemistry becomes more complex, and reactions cannot be easily generalized in a condensed mechanism. Numerous unsaturated compounds are emitted but SAPRC07TB only represents ethene and propene explicitly; the rest are grouped into two model species, OLE1 and OLE2, depending on reaction with OH. On a regional basis, ethene, propene and methylsubstituted butenes make up a majority of alkene emissions (Fig. 8). Reactions of these alkenes with OH e the most important decay pathway e have estimated uncertainties of 20e30% (Calvert et al., 2000). In these reactions, the description of b-hydroxy peroxy and alkoxy radical reactions that follow the initial reactions have either been measured or estimated, and explicitly representing glycoaldehyde in SAPRC07TB allows for accurate representation of the fate of b-hydroxyalkoxy radicals from ethene. For larger alkenes, isomerization and reaction with oxygen play a larger role. Reactions of alkenes with ozone are less straightforward - reactions of Criegee intermediates and OH radical generation is an area requiring additional laboratory and modeling studies (Calvert et al., 2000). A 30% uncertainty in chemistry could change aldehyde predictions by 3e24%, although concentrations could increase or decrease. 4.5. Alkanes Methane and other alkanes contribute to HCHO and CH3CHO formation. Simulations using the CHIMERE model (Dufour et al., 2009) also identified methane as the dominant source of HCHO in Western Europe, with isoprene playing a lesser role. Despite the low reactivity of methane (maximum incremental reactivity for ozone 0.001 times that of isoprene) and exclusion from the definition of VOC in the United States (US GPO, 2011), large emissions make methane an important hemispheric driver of atmospheric photochemistry. Methane contributes to ozone formation (Fiore et al., 2008; Jeffries, 1995); our results finds methane is also an important contributor to HCHO. Larger alkanes can contribute up to 20% of CH3CHO formation.

D.J. Luecken et al. / Atmospheric Environment 47 (2012) 477e490

487

Fig. 8. Annual anthropogenic emissions of 15 alkenes with the largest emission rates, and their allocation by source categories in the NEI. The allocation of explicit species to SAPRC07 model species (OLE1 and OLE2) is also shown.

Anthropogenic sources of alkanes generally dominate biogenic sources in our simulations. Fig. 9 shows annual emissions of methane and the fifteen largest VOC alkanes. Mobile sources are important, along with other industrial, waste and transport emissions. Landfills are a major source of methane, while storage and transport of petroleum contributes to n-butane emissions. Anthropogenic emissions from the NEI have been reported to differ from emission ratios in the Northeastern US by up to an order of magnitude (Warneke et al., 2007) based on measurements and calculated photochemical age. These discrepancies with the NEI must be further examined to quantify uncertainty in alkane emissions. Alkane chemistry is generally better understood than other species, although the representation of larger and multifunctional alkanes in air quality models is highly condensed. Despite uncertainties inherent in representing a large number of species by a small subset, Derwent et al. (2010) found that the POCP of alkanes modeled with SAPRC07 correlated well with those using MCM, a more detailed representation. This gives confidence that SAPRC07 represents our knowledge of alkane degradation similarly to MCM. 4.6. Ethanol Potential changes in acetaldehyde concentrations from fuel reformulation has been recently studied (Cook et al., 2011; Jacobson, 2007). Our simulations predict ethanol contributes only

3e6% of photochemically-produced acetaldehyde. One recent study (Millet et al., 2010) calculated that converting the US vehicle fleet to 85% ethanol fuel would increase acetaldehyde emissions by 21% and double ethanol emissions. We estimate that such a change would increase acetaldehyde by an average of 6%, possibly more near roadways. Using ethanol fuels will affect concentrations of other VOCs (Cook et al., 2011; Graham et al., 2008; Jacobson, 2007; US EPA, 2010), which could be as important as changes in CH3CHO or ethanol. Changes in alkene emissions, in particular, would impact acetaldehyde. Effects of fuel reformulations on these other VOCs must be accurately characterized to predict their impact on acetaldehyde. 4.7. Other chemicals Our analysis does not find a major role for aromatic compounds in producing HCHO or CH3CHO. While aromatic chemistry is still uncertain, even a major increase in HCHO formation from aromatics would not make them significant compared to other chemicals. Emissions of methanol and acetone (part of the lumped oxygenates category) contribute less than 2% of total HCHO. Recent analyses of aircraft data in the Pacific Northwest (Xie et al., 2011a) postulate that emissions of methanol and acetone would have to increase by factors of 30 and 25, respectively, to account for observed values of these compounds. If acetone emissions were 25

488

D.J. Luecken et al. / Atmospheric Environment 47 (2012) 477e490

Fig. 9. Annual anthropogenic emissions of methane and the highest 15 non-methane alkanes and their allocation by source categories in the NEI. The allocation of explicit species to SAPRC07 model species (ALK1 through ALK5) is also shown.

times too low in the domain modeled here, HCHO could increase by nearly 20%, bringing the modeled values closer to observations. Increasing methanol would increase its role in HCHO formation similarly, but isoprene would continue to remain the most important precursor in most areas in summer. Neither methanol nor acetone is considered a VOC (US GPO, 2011) and must be derived from inventories of total organic gas. Lastly, is it possible that our simulations underestimate OH radical concentrations, thus underestimating a critical first step in degradation of stable VOCs? Studies comparing box model predictions to measurements find both underestimates (Chen et al., 2010; Hofzumahaus et al., 2009) and some overestimates (Emmerson et al., 2007). Measurements of OH and HO2 are difficult so mechanisms have not been thoroughly evaluated for predictions of these radicals. OH participates in both HCHO production and decay, so errors in OH will not change HCHO in a linear fashion. If OH is larger than predicted by SAPRC07TB, this would explain some of the HCHO and CH3CHO underpredictions. 5. Summary Because the chemistry of HCHO and CH3CHO is integral to almost all VOC decay and production pathways, it reflects our current understanding of VOC photochemistry and tests our entire mechanistic understanding of atmospheric chemistry. To accurately model their concentrations, we must correctly account for

emissions, transport, deposition, and the degradation chemistry of not only aldehydes, but also VOC precursors to aldehydes. In this study we conclude that: 1. Atmospheric photochemical production accounts for most of the HCHO and CH3CHO concentrations in the ambient atmosphere. Aside from dense urban/industrial areas, reduction of HCHO and CH3CHO emissions is expected to have a small effect on reducing concentrations. In a few areas, especially in winter, emissions are important and a large source of uncertainty is differences between emissions from the NEI and emissions based on speciated VOC. 2. Substantial spatial and seasonal variation occurs in how much various VOCs contribute to photochemical production of aldehydes. For example, isoprene is not always the dominant VOC producing HCHO, even during summer. 3. Biogenic emissions of compounds that photochemically react to produce HCHO and CH3CHO are highly uncertain and a potential source of error in aldehyde predictions. 4. Resolving uncertainties in the chemistry of isoprene and other alkenes is critical for chemical mechanisms to accurately predict concentrations of aldehydes. Newly proposed isoprene mechanisms must be evaluated for a wide range of species, including HCHO. 5. Anthropogenic alkenes, largely from mobile sources, play an important role in aldehyde formation. While mobile sources

D.J. Luecken et al. / Atmospheric Environment 47 (2012) 477e490

are perhaps better represented in the NEI than other categories, new uncertainties are constantly being introduced by changes in vehicles and fuels. 6. The CMAQ air quality model tends to underpredict aldehyde concentrations. At many sites this underprediction is close to 50%. While much effort has spent developing condensed chemical mechanisms that are as representative of the underlying chemicals as possible, there will necessarily be compromises. Even the most explicit mechanisms, such as the MCM, rely on some degree of condensation. It is encouraging that comparisons between explicit mechanisms and condensed mechanisms are generally consistent (Archibald et al., 2010b; Derwent et al., 2010), although carbonyls tend to be one class of compounds which are often predicted differently among mechanisms. The general underprediction of air quality models for aldehydes is an area which requires more investigation. Because aldehyde decay is a large source of new radicals which drive photochemistry, this raises questions about the robustness of predictions for other photochemical pollutants, such as ozone and PM2.5; species which air quality models predict well. HCHO, in particular, is an important ozone precursor. Our findings on the most significant chemical sources of HCHO and CH3CHO are based on our best assessment of emissions and chemistry on a regional scale. The existence of many uncertainties implies that the contributions of different VOCs to aldehyde formation may change somewhat as our knowledge of chemistry and emissions improves. Refining our understanding of aldehyde behavior in the atmosphere is an iterative and imperfect process involving measurements, modeling, and collecting improved data on emission sources. New information will help to more accurately bound our understanding of how air quality might change under future emissions scenarios, and build confidence in using regional models for air quality management and research. Acknowledgments The authors acknowledge helpful discussions with our US EPA colleagues Ted Palma and Ying Xie, and with William Carter, who developed this SAPRC07TB mechanism. We greatly appreciate the numerous helpful suggestions from our internal and external reviewers, all of whom have helped focused the purpose of this manuscript and clarified the presentation. The United States Environmental Protection Agency, through its Office of Research and Development, funded and managed the research described here. It has been subjected to Agency administrative review and approved for publication. Appendix. Supplementary data Supplementary data associated with this article can be found in the online version, at doi:10.1016/j.atmosenv.2011.10.005. References Adewuyi, Y.G., Cho, S.Y., Tsay, R.P., Carmichael, G.R., 1984. Importance of formaldehyde in cloud chemistry. Atmospheric Environment 18, 2413e2420. Archibald, A.T., Cooke, M.C., Utembe, S.R., Shallcross, D.E., Derwent, R.G., Jenkin, M.E., 2010a. Impacts of mechanistic changes on HOx formation and recycling in the oxidation of isoprene. Atmospheric Chemistry Physics 10, 8097e8118. Archibald, A.T., Jenkin, M.E., Shallcross, D.E., 2010b. An isoprene mechanism intercomparison. Atmospheric Environment 44, 5356e5364. Arnts, R.R., Tejada, S.B., 1989. 2,4-Dinitrophenylhydrazine-coated silica gel cartridge method for determination of formaldehyde in air: identification of an ozone interference. Environmental Science & Technology 23, 1428e1430.

489

Atkinson, R., 1997. Gas-phase tropospheric chemistry of volatile organic compounds: 1. alkanes and alkenes. Journal of Physical and Chemical Reference Data 26, 215e290. Byun, D., Schere, K.L., 2006. Review of the governing equations, computational algorithms, and other components of the models-3 community multiscale air quality (CMAQ) modeling system. Applied Mechanics Reviews 59, 51e77. Cai, C., Kelly, J.T., Avise, J.C., Kaduwela, A.P., Stockwell, W.R., 2011. Implementation of the SAPRC07C chemical mechanism in the Community Multiscale Air Quality (CMAQ) model for photochemical modeling in California: Comparisons with SAPRC99. Journal of the Air & Waste Management Association 61, 559e572. Calvert, J.G., Atkinson, R., Kerr, J.A., Madronich, S., Moortgat, G.K., Wallington, T.J., Yarwood, G., 2000. The Mechanisms of Atmospheric Oxidation of the Alkenes. Oxford University Press, New York/Oxford. Carslaw, N., Bell, N., Lewis, A.C., McQuaid, J.B., Pilling, M.J., 2000. A detailed case study of isoprene chemistry during the EASE96 Mace Head campaign. Atmospheric Environment 34, 2827e2836. Carter, W.P.L., 2010. Development of the SAPRC-07 chemical mechanism. Atmospheric Environment 44, 5324e5335. Chen, S., Ren, X., Mao, J., Chen, Z., Brune, W.H., Lefer, B., Rappenglück, B., Flynn, J., Olson, J., Crawford, J.H., 2010. A comparison of chemical mechanisms based on TRAMP-2006 field data. Atmospheric Environment 44, 4116e4125. Choi, W., Faloona, I.C., Bouvier-Brown, N.C., McKay, M., Goldstein, A.H., Mao, J., Brune, W.H., LaFranchi, B.W., Cohen, R.C., Wolfe, G.M., Thornton, J.A., Sonnenfroh, D.M., Millet, D.B., 2010. Observations of elevated formaldehyde over a forest canopy suggest missing sources from rapid oxidation of arboreal hydrocarbons. Atmospheric Chemistry Physics 10, 8761e8781. Cook, R., Glover, E., Yanca, C., Warila, J., Sales, A., 2010. Modeling Air Toxics in EPA’s MOVES Emissions Model, 2010 Coordinating Research Council Air Toxics Workshop. Coordinating Research Council, Sacramento, CA. Cook, R., Phillips, S., Houyoux, M., Dolwick, P., Mason, R., Yanca, C., Zawacki, M., Davidson, K., Michaels, H., Harvey, C., Somers, J., Luecken, D., 2011. Air quality impacts of increased use of ethanol under the United States’ Energy Independence and Security Act. Atmospheric Environment 45, 7714e7724. Derwent, R.G., Jenkin, M.E., Pilling, M.J., Carter, W.P.L., Kaduwela, A., 2010. Reactivity scales as comparative tools for chemical mechanisms. Journal of the Air & Waste Management Association 60, 914e924. Doraiswamy, P., Hogrefe, C., Hao, W., Henry, R.F., Civerolo, K., Ku, J.Y., Sistla, G., Schwab, J.J., Demerjian, K.L., 2009. A diagnostic comparison of measured and model-predicted speciated VOC concentrations. Atmospheric Environment 43, 5759e5770. Dufour, G., Wittrock, F., Camredon, M., Beekmann, M., Richter, A., Aumont, B., Burrows, J.P., 2009. SCIAMACHY formaldehyde observations: constraint for isoprene emission estimates over Europe? Atmospheric Chemistry Physics 9, 1647e1664. Emmerson, K.M., Carslaw, N., Carslaw, D.C., Lee, J.D., McFiggans, G., Bloss, W.J., Gravestock, T., Heard, D.E., Hopkins, J., Ingham, T., Pilling, M.J., Smith, S.C., Jacob, M., Monks, P.S., 2007. Free radical modelling studies during the UK TORCH Campaign in Summer 2003. Atmospheric Chemistry Physics 7, 167e181. Fiore, A.M., West, J.J., Horowitz, L.W., Naik, V., Schwarzkopf, M.D., 2008. Characterizing the tropospheric ozone response to methane emission controls and the benefits to climate and air quality. Journal of Geophysical Research 113, D08307. Foley, K.M., Roselle, S.J., Appel, K.W., Bhave, P.V., Pleim, J.E., Otte, T.L., Mathur, R., Sarwar, G., Young, J.O., Gilliam, R.C., Nolte, C.G., Kelly, J.T., Gilliland, A.B., Bash, J.O., 2009. Incremental testing of the community multiscale air quality (CMAQ) modeling system version 4.7. Geoscientific Model Development Discussions 2, 1245e1297. Fried, A., Walega, J.G., Olson, J.R., Crawford, J.H., Chen, G., Weibring, P., Richter, D., Roller, C., Tittel, F.K., Heikes, B.G., Snow, J.A., Shen, H., O’Sullivan, D.W., Porter, M., Fuelberg, H., Halland, J., Millet, D.B., 2008. Formaldehyde over North America and the North Atlantic during the summer 2004 INTEX campaign: methods, observed distributions, and measurement-model comparisons. Journal of Geophysical Research 113, D10302. Friedfeld, S., Fraser, M., Ensor, K., Tribble, S., Rehle, D., Leleux, D., Tittel, F., 2002. Statistical analysis of primary and secondary atmospheric formaldehyde. Atmospheric Environment 36, 4767e4775. Garcia, A.R., Volkamer, R., Molina, L.T., Molina, M.J., Samuelson, J., Mellqvist, J., Galle, B., Herndon, S.C., Kolb, C.E., 2006. Separation of emitted and photochemical formaldehyde in Mexico City using a statistical analysis and a new pair of gas-phase tracers. Atmospheric Chemistry Physics 6, 4545e4557. Gipson, G.L., 1999. Process analysis. In: Ching, J.K.S. (Ed.), Science Algorithms of the EPA Models-3 Community Multiscale Air Quality (CMAQ) Modeling System. U.S. EPA, Washington, DC, pp. 16-11e16-39. Graham, L.A., Belisle, S.L., Baas, C.L., 2008. Emissions from light duty gasoline vehicles operating on low blend ethanol gasoline and E85. Atmospheric Environment 42, 4498e4516. Grosjean, D., Swanson, R.D., Ellis, C., 1983. Carbonyls in Los Angeles air: contribution of direct emissions and photochemistry. The Science of the Total Environment 29, 65e85. Guenther, A., Karl, T., Harley, P., Wiedinmyer, C., Palmer, P.I., Geron, C., 2006. Estimates of global terrestrial isoprene emissions using MEGAN (Model of Emissions of Gases and Aerosols from Nature). Atmospheric Chemistry Physics 6, 3181e3210. HEI, 2007. In: Panel, H.A.T.R. (Ed.), Mobile-Source Air Toxics: A Critical Review of the Literature on Exposure and Health Effects. Health Effects Institute, Boston, Massachusetts.

490

D.J. Luecken et al. / Atmospheric Environment 47 (2012) 477e490

Herrington, J.S., Fan, Z.H., Lioy, P.J., Zhang, J., 2006. Low acetaldehyde collection efficiencies for 24-hour sampling with 2,4-dinitrophenylhydrazine (DNPH)coated solid sorbents. Environmental Science & Technology 41, 580e585. Hofzumahaus, A., Rohrer, F., Lu, K., Bohn, B., Brauers, T., Chang, C.C., Fuchs, H., Holland, F., Kita, K., Kondo, Y., Li, X., Lou, S., Shao, M., Zeng, L., Wahner, A., Zhang, Y., 2009. Amplified trace gas removal in the troposphere. Science 324, 1702e1704. Hogrefe, C., Isukapalli, S.S., Tang, X., Georgopoulos, P., He, S., Zalewsky, E.E., Hao, W., Ku, J.Y., Key, T., Sistla, G., 2011. Impact of biogenic emission uncertainties on the simulated response of ozone and fine particulate matter to anthropogenic emission reductions. Journal of the Air & Waste Management Association 61, 92e108. Houyoux, M., Strum, M., Mason, R., Dolwick, P., Michaels, H., Huang, A., Allen, C.T., Beidler, J., 2009. Implementation of MOVES-based PM2.5 emissions approach for onroad gasoline sources using hourly, gridded temperatures. 18th Annual International Emission Inventory Conference, Baltimore, MD. Hutzell, W.T., Luecken, D.J., Appel, K.W., Carter, W.P.L., 2012. Interpreting predictions from the SAPRC07 mechanism based on regional and continental simulations. Atmospheric Environment 46, 417e429. Jacobson, M.Z., 2007. Effects of Ethanol (E85) versus gasoline vehicles on cancer and mortality in the United States. Environmental Science & Technology 41, 4150e4157. Jang, J.C.C., Jeffries, H.E., Tonnesen, S., 1995. Sensitivity of ozone to model grid resolution e II. Detailed process analysis for ozone chemistry. Atmospheric Environment 29, 3101e3114. Jeffries, H.E., 1995. Photochemical air pollution. In: Singh, H.B. (Ed.), Composition, Chemistry, and Climate of the Atmosphere. Van Nostrand Reinhold, New York. Jeffries, H.E., Tonnesen, S., 1994. A comparison of two photochemical reaction mechanisms using mass balance and process analysis. Atmospheric Environment 28, 2991e3003. Karst, U., Binding, N., Cammann, K., Witting, U., 1993. Interferences of nitrogen dioxide in the determination of aldehydes and ketones by sampling on 2,4dinitrophenylhydrazine-coated solid sorbent. Fresenius’ Journal of Analytical Chemistry 345, 48e52. Kim, S., Guenther, A., Karl, T., Greenberg, J., 2011. Branch-level measurement of total OH reactivity for constraining unknown BVOC emission during the CABINEX (Community Atmosphere-Biosphere INteractions Experiments)-09 Field Campaign. Atmospheric Chemistry Physics Discussions 11, 7781e7809. Kite, C., 2011. Preliminary comparison between MOVES and MOBILE6 Houston/ Galveston/Brazoria (HGB) on-road emission inventories for 2006 and 2018. Presented to the Regional Air Quality Planning Committee, Houston, TX. Luecken, D.J., Hutzell, W.T., Gipson, G.L., 2006. Development and analysis of air quality modeling simulations for hazardous air pollutants. Atmospheric Environment 40, 5087e5096. Millet, D.B., Guenther, A., Siegel, D.A., Nelson, N.B., Singh, H.B., de Gouw, J.A., Warneke, C., Williams, J., Eerdekens, G., Sinha, V., Karl, T., Flocke, F., Apel, E., Riemer, D.D., Palmer, P.I., Barkley, M., 2010. Global atmospheric budget of acetaldehyde: 3-D model analysis and constraints from in-situ and satellite observations. Atmospheric Chemistry Physics 10, 3405e3425. Millet, D.B., Jacob, D.J., Boersma, K.F., Fu, T.M., Kurosu, T.P., Chance, K., Heald, C.L., Guenther, A., 2008. Spatial distribution of isoprene emissions from North America derived from formaldehyde column measurements by the OMI satellite sensor. Journal of Geophysical Research 113, D02307. Millet, D.B., Jacob, D.J., Turquety, S., Hudman, R.C., Wu, S., Fried, A., Walega, J., Heikes, B.G., Blake, D.R., Singh, H.B., Anderson, B.E., Clarke, A.D., 2006. Formaldehyde distribution over North America: implications for satellite retrievals of formaldehyde columns and isoprene emission. Journal of Geophysical Research 111, D24S02. Palmer, P.I., Abbot, D.S., Fu, T.M., Jacob, D.J., Chance, K., Kurosu, T.P., Guenther, A., Wiedinmyer, C., Stanton, J.C., Pilling, M.J., Pressley, S.N., Lamb, B., Sumner, A.L., 2006. Quantifying the seasonal and interannual variability of North American isoprene emissions using satellite observations of the formaldehyde column. Journal of Geophysical Research 111, D12315. Paulot, F., Crounse, J.D., Kjaergaard, H.G., Kroll, J.H., Seinfeld, J.H., Wennberg, P.O., 2009. Isoprene photooxidation: new insights into the production of acids and organic nitrates. Atmospheric Chemistry Physics 9, 1479e1501. Peeters, J., 2010. HOx radical regeneration in isoprene oxidation via peroxy radical isomerisations. II: experimental evidence and global impact. Physical Chemistry Chemical Physics: PCCP 12, 14227. Rappenglück, B., Dasgupta, P.K., Leuchner, M., Li, Q., Luke, W., 2010. Formaldehyde and its relation to CO, PAN, and SO2 in the HoustoneGalveston airshed. Atmospheric Chemistry Physics 10, 2413e2424.

Ren, X., Brune, W.H., Mao, J., Mitchell, M.J., Lesher, R.L., Simpas, J.B., Metcalf, A.R., Schwab, J.J., Cai, C., Li, Y., 2006. Behavior of OH and HO2 in the winter atmosphere in New York City. Atmospheric Environment 40, 252e263. Saunders, S.M., Jenkin, M.E., Derwent, R.G., Pilling, M.J., 2003. Protocol for the development of the Master Chemical Mechanism, MCM v3 (Part A): tropospheric degradation of non-aromatic volatile organic compounds. Atmospheric Chemistry Physics 3, 161e180. Seinfeld, J.H., Pandis, S.N., 1998. Atmospheric Chemistry and Physics. John Wiley & Sons, Inc., New York. Simon, H., Beck, L., Bhave, P., Divita, F., Hsu, Y., Luecken, D., Mobley, D., Pouliot, G., Reff, A., Sarwar, G., Strum, M., 2010. The development and uses of EPA’s SPECIATE database. Atmospheric Pollution Research 1, 196e206. Sirju, A.P., Shepson, P.B., 1995. Laboratory and field investigation of the DNPH Cartridge Technique for the measurement of atmospheric carbonyl compounds. Environmental Science & Technology 29, 384e392. Stavrakou, T., Peeters, J., Müller, J.F., 2010. Improved global modelling of HOx recycling in isoprene oxidation: evaluation against the GABRIEL and INTEX-A aircraft campaign measurements. Atmospheric Chemistry Physics 10, 9863e9878. Steiner, A.L., Cohen, R.C., Harley, R.A., Tonse, S., Millet, D.B., Schade, G.W., Goldstein, A.H., 2008. VOC reactivity in central California: comparing an air quality model to ground-based measurements. Atmospheric Chemistry Physics 8, 351e368. Sumner, A.L., Shepson, P.B., Couch, T.L., Thornberry, T., Carroll, M.A., Sillman, S., Pippin, M., Bertman, S., Tan, D., Faloona, I., Brune, W., Young, V., Cooper, O., Moody, J., Stockwell, W., 2001. A study of formaldehyde chemistry above a forest canopy. Journal of Geophysical Research 106, 24387e24405. Tilgner, A., Herrmann, H., 2010. Radical-driven carbonyl-to-acid conversion and acid degradation in tropospheric aqueous systems studied by CAPRAM. Atmospheric Environment 44, 5415e5422. US EPA, 2010. Renewable Fuel Standard Program (RFS2) Regulatory Impact Analysis, February, 2010 ed. U.S Environmental Protection Agency. US EPA, 2011. 2005 National-Scale Air Toxics Assessment. U.S. Environmental Protection Agency. US GPO, 2011. Code of Federal Requirements, Title 40: Protection of the Environment. US Government Printing Office. Warneck, P., 1989. Sulfur dioxide in rain clouds: gaseliquid scavenging efficiencies and wet deposition rates in the presence of formaldehyde. Journal of Atmospheric Chemistry 8, 99e117. Warneke, C., de Gouw, J.A., Del Negro, L., Brioude, J., McKeen, S., Stark, H., Kuster, W.C., Goldan, P.D., Trainer, M., Fehsenfeld, F.C., Wiedinmyer, C., Guenther, A.B., Hansel, A., Wisthaler, A., Atlas, E., Holloway, J.S., Ryerson, T.B., Peischl, J., Huey, L.G., Hanks, A.T.C., 2010. Biogenic emission measurement and inventories determination of biogenic emissions in the eastern United States and Texas and comparison with biogenic emission inventories. Journal of Geophysical Research 115, D00F18. Warneke, C., McKeen, S.A., de Gouw, J.A., Goldan, P.D., Kuster, W.C., Holloway, J.S., Williams, E.J., Lerner, B.M., Parrish, D.D., Trainer, M., Fehsenfeld, F.C., Kato, S., Atlas, E.L., Baker, A., Blake, D.R., 2007. Determination of urban volatile organic compound emission ratios and comparison with an emissions database. Journal of Geophysical Research 112, D10S47. Wert, B.P., Trainer, M., Fried, A., Ryerson, T.B., Henry, B., Potter, W., Angevine, W.M., Atlas, E., Donnelly, S.G., Fehsenfeld, F.C., Frost, G.J., Goldan, P.D., Hansel, A., Holloway, J.S., Hubler, G., Kuster, W.C., Nicks Jr., D.K., Neuman, J.A., Parrish, D.D., Schauffler, S., Stutz, J., Sueper, D.T., Wiedinmyer, C., Wisthaler, A., 2003. Signatures of terminal alkene oxidation in airborne formaldehyde measurements during TexAQS 2000. Journal of Geophysical Research 108, 4104. Wiedinmyer, C., Greenberg, J., Guenther, A., Hopkins, B., Baker, K., Geron, C., Palmer, P.I., Long, B.P., Turner, J.R., Pétron, G., Harley, P., Pierce, T.E., Lamb, B., Westberg, H., Baugh, W., Koerber, M., Janssen, M., 2005. Ozarks Isoprene Experiment (OZIE): measurements and modeling of the “isoprene volcano”. Journal of Geophysical Research 110, D18307. Xie, Y., Elleman, R., Jobson, T., Lamb, B, 2011a. Evaluation of O3-NOx-VOC sensitivities predicted with the CMAQ photochemical model using Pacific Northwest 2001 field observations. Journal of Geophysical Research 116, D20303. Xie, Y., Paulot, F., Pinder, R.W., Carter, W.P.L., Nolte, C.G., Luecken, D.J., Hutzell, W.T., Wennberg, P.O., Cohen, R., 2011b. Understanding the impact of isoprene nitrates and OH reformation on regional air quality using recent advances in isoprene photooxidation chemistry. In: 9th Annual CMAS Conference, Chapel Hill, NC.